SiteLens MCP · Browser Intelligence in the IDE

Automate and inspect real browsers locally. No account. No API key. Free.

SiteLens Local runs Playwright on your machine through @sitelensapi/mcp. Use it from Cursor or Claude Code to open localhost, click and type, save screenshots, read console and network failures, and export a PDF report. Cloud tools are optional and need an API key. The published package does not send named capture targets.

npm install Browser Intelligence DevTools guide Cloud/API MCP + keys
Local browser QA from your editor — SiteLens Local via MCP
Cloud/API run digests — optional MCP resources when enabled

SiteLens Local via MCP

Free local execution

Install @sitelensapi/mcp from npm (any project—no SiteLens monorepo required), run sitelens_setup, then Local tools against http://localhost:…—Playwright-powered Chromium, Firefox, and WebKit on your machine (plus installed Chrome/Edge channels). Beyond smoke QA: session DevTools tools, diagnostics, sitelens_compare_browser_runs, unified timeline, and Playwright Codegen for authoring reusable .sitelens/flows. Saved auth under ~/.sitelens; flows under .sitelens/flows. See Browser Intelligence and the DevTools guide.

Geographic capture from MCP

Paid · SiteLens Pro

See the web from anywhere. sitelens_capture accepts location.country on cloud POST /capture. Country exit is SiteLens Pro (capture.geographic). Local MCP runs stay free and do not route through a geographic proxy. sitelens_list_capture_countries reads the public country list and does not capture.

MCP geographic capture guide · Product page

Agents can help author automation — not just run it

Playwright Codegen

An AI coding agent can participate in the automation lifecycle: start a headed Playwright Codegen session, inspect status, finalize or cancel, convert recorded behavior into a SiteLens flow, then run and inspect evidence through the existing SiteLens workflow.

AgentRecord / createSiteLens flowRunBrowser IntelligenceAgent

Codegen is an authoring input. SiteLens converts Playwright-generated behavior into reusable native flows under .sitelens/flows/*.json — the same format Desktop and CLI use. Conversion is loss-aware (complete vs partial); SiteLens never silently treats unmapped behavior as success. Partial saves require an explicit opt-in.

MCP lifecycle tools

  • sitelens_flow_codegen_start
  • sitelens_flow_codegen_status
  • sitelens_flow_codegen_finalize
  • sitelens_flow_codegen_cancel
  • sitelens_flow_codegen_convert — paste / convert source

CLI & Desktop

Same conversion path from the terminal or the macOS app:

sitelens flow codegen --flow-id login-recorded \
  --url http://localhost:3000/login \
  --browser chromium --save

In SiteLens Desktop, use Record Flow in Automation.

Headed recording still needs a human (or local operator) in the browser window. Agents orchestrate the session and conversion — they do not magically perform the clicks inside an MCP client chat alone.

npm install

Published

Run in any terminal before wiring Cursor. Package: @sitelensapi/mcp.

npm install @sitelensapi/mcp
npx @sitelensapi/mcp sitelens-mcp setup
npx @sitelensapi/mcp sitelens-mcp -- --self-test
npx @sitelensapi/mcp sitelens doctor

setup installs the default Chromium runtime (no Docker); Local QA also supports Firefox, WebKit, and Chrome/Edge channels. --self-test and doctor verify runtime readiness.

Start in Cursor or Claude Code

stdio · local

Same server for both. SITELENS_MODE=local registers local tools only. Do not put setup or doctor in args — those exit instead of serving MCP.

Cursor

~/.cursor/mcp.json or the project .cursor/mcp.json. Restart Cursor after saving.

{
  "mcpServers": {
    "sitelens": {
      "command": "npx",
      "args": ["-y", "@sitelensapi/mcp"],
      "env": {
        "SITELENS_MODE": "local",
        "SITELENS_DEFAULT_URL": "http://127.0.0.1:3000"
      }
    }
  }
}

Claude Code

Project .mcp.json, same stdio command. No SiteLens account.

{
  "mcpServers": {
    "sitelens": {
      "command": "npx",
      "args": ["-y", "@sitelensapi/mcp"],
      "env": { "SITELENS_MODE": "local" }
    }
  }
}

A localhost pass

Ask the agent to use these published tools, in order. Ad-hoc QA accepts up to 8 steps. Repo flows go further.

  1. sitelens_setup — install the default Chromium runtime.
  2. sitelens_run_local_qa on http://127.0.0.1:3000 — load the app, run the smoke steps, and keep a screenshot.
  3. sitelens_session_start, then sitelens_session_run_steps with goto, click, and fill when you need a specific path. sitelens_session_close when finished.
  4. sitelens_get_console_messages and sitelens_get_network_requests on the localRunId when something fails.
  5. sitelens_report_generate, then sitelens_report_export_pdf — a PDF of the local report, not a PDF of the webpage.

Authenticated apps use sitelens_auth_start and sitelens_auth_save on your machine. See localhost auth QA and Cursor setup.

Use SiteLens Local with Cursor

MCP

Configure SiteLens MCP in Cursor (sitelens_setup, sitelens_run_local_qa, sitelens_run_qa, …): Local tools run Playwright browsers on your machine with no key (Chromium by default; Firefox, WebKit, and Chrome/Edge channels supported); Cloud/API tools call hosted execution when you set SITELENS_API_KEY.

Five-minute MCP workflow

Local-first: run npm install @sitelensapi/mcp and npx @sitelensapi/mcp sitelens-mcp setup, paste MCP config → sitelens_setup → SiteLens Local on localhost. Add dashboard sign-up and an API key only when you want SiteLens Cloud/API runs and persisted history.

1
Install from npm 
npm install @sitelensapi/mcp
npx @sitelensapi/mcp sitelens-mcp setup
2
Paste MCP config

Into ~/.cursor/mcp.json · restart Cursor. See dashboard MCP cockpit for a copy-paste JSON block.

3
Run setup 
Use the SiteLens MCP tool sitelens_setup.

Installs the default Chromium runtime and checks artifact dirs—no API key required for Local tools. Local QA also supports Firefox, WebKit, and Chrome/Edge channels.

4
SiteLens Local run

sitelens_run_local_qa for localhost and private URLs—runs on your machine at no cost.

5
Optional: SiteLens Cloud/API account

Sign up in the dashboard when you need hosted public capture, cloud QA, or persisted cloud history. Local runs stay on your machine.

6
Optional: API key + Cloud/API QA

Set SITELENS_API_KEY, then sitelens_run_qa for public URLs and persisted cloud runs.

MCP-ready prompts

Cursor

Each line maps to a concrete MCP tool—adjust URLs or flow ids to match your app.

Bootstrap

Use the SiteLens MCP tool sitelens_setup.

SiteLens Local smoke

Use the SiteLens MCP tool sitelens_run_local_qa to QA http://localhost:3000 desktop + mobile.

Compare Local runs

Use the SiteLens MCP tool sitelens_compare_local_runs to contrast my latest two SiteLens Local bundles.

SiteLens Cloud/API pass

Use the SiteLens MCP tool sitelens_run_qa against https://example.com with desktop + mobile viewports.

Cloud/API run memory (optional)

When you use Cloud/API tools, MCP can surface durable ids (sitelens_list_runs, sitelens_get_run) and resources such as sitelens://runs/{runId} for regressions across chat sessions—alongside persisted SiteLens Cloud/API history and SiteLens Studio flow runs.

Scenario Runner chains reusable flows in a shared browser session—saved auth profiles, parameterized flows, and multi-step workflow verification via sitelens_run_local_scenario. See the Scenario Runner guide.

Setup guides (npm README)

Public docs for @sitelensapi/mcp — no private repo links.

Related

Contact & onboarding

Beta access, MCP rollout help, or product updates — same short form everywhere.